research data management 101: the lifecycle of a...
Post on 19-Aug-2018
214 Views
Preview:
TRANSCRIPT
Research Data Management 101:The Lifecycle of a Dataset
IAPJanuary 12, 2016
Data Management Services @ MIT Libraries
Contact: data-management@mit.edu
• Workshops
• Web guide: http://libraries.mit.edu/data-management
• Individual consultations includes help with creating data management plans
guide you to related services across MIT
Today’s session: topics
File organization
Today’s session: the research data lifecycle
File organization
Study Concept
Data Collection
Data Processing
Data Distribution
Data Analysis
Data Discovery
Data ArchivingRepurposing
Lifecycle adapted from DDI version 3.0 Combined Life Cycle Model
Research data lifecycle phases
Study Concept
Data Collection
Data Processing
Data Distribution
Data Analysis
Data Discovery
Data ArchivingRepurposing
Lifecycle adapted from DDI version 3.0 Combined Life Cycle Model
Pre-project
Active phasePost-project
Secondary use
What do we mean by data?
General Social Sciences Natural/PhysicalSciences
• images
• video
• mapping/GIS data
• numerical measurements
• survey responses
• focus group and individual interviewtranscripts
• economic indicators
• demographics
• opinion polling
• measurements generated by sensors/laboratory instruments
• computer modeling
• simulations
• observations and/or field studies
• specimen
A case study
Pre-project: planning
Study Concept
Data Collection
Data Processing
Data Distribution
Data Analysis
Data Discovery
Data ArchivingRepurposing
Lifecycle adapted from DDI version 3.0 Combined Life Cycle Model
Where do you start?
1. Consider your goals.
What do you want to get out of managing your data?
• Part of ongoing research• Compliance with funder or publisher requirements:• Dissemination for others’ use
Where do you start?
1. Consider your goals.
2. What are you collecting?
3. What are you keeping?
Where do you start?
1. Consider your goals.
2. What are you collecting?
3. What are you keeping?
4. Where do you want to keep it?
Where do you start?
1. Consider your goals.
2. What are you collecting?
3. What are you keeping?
4. Where do you want to keep it?
5. What do you need to be able to use it & share it later?
Funder requirements
• Funding agencies increasingly requiring people to share outputs of research, including data (and publications)
• Their motivation: extend the impact of their research farther• What funders require data sharing?
• Most federal agencies and many private funders as well• What do funders require?
• Data management plan (DMP) submitted with grant application
• Can request funds for data management in the grant• Sharing of final data produced
• See: http://libraries.mit.edu/oarequirements• In our case study, my research is funded by NSF; I have to:
• Share all of the data generated in my project• Include a DMP in my grant application
What can help me write my DMP? DMPTool!
• New tool to help you create your data management plan• Guides you through writing a DMP specific to a funder’s
requirements using templates• Access MIT-specific information and resources related to
data management• Request feedback on your DMP• Access the tool at http://dmptool.org/ and log in
using Touchstone
Data Documentation(AKA Metadata)
Metadata: why does it matter?
Data is not self-describing.
Metadata, or “data about data” explains your dataset and allows you to document important information for:
• Finding the data later• Understanding what the data is later• Sharing the data (both with collaborators and future
secondary data users)• Consider it an investment of time that will save you
trouble later several-fold
Metadata standards
Examples:• FGDC (Federal Geographic Data Committee)• DDI (Data Documentation Initiative)• Dublin Core• Darwin Core• ABCD (Access to Biological Collections Data)• AVMS (Astronomy Visualization Metadata Standard)• CSDGM (Content Standard for Digital Geospatial Metadata)
Advantages:• Ensure you have a complete, standard set of information about each
part of your data• Enable your dataset to be organized with other datasets
Metadata
Do what works for you!
Document and describe your data in whatever way works for you.
Better “good enough” than doing nothing
Possible metadata options:
1. Dublin Core• General metadata
standard• Widely applicable• Used in many different
repositories
2. Darwin Core• For biological diversity• Emphasizes taxonomy,
which I don’t care about• Frequently used in
biodiversity databases
Metadata: our case study
Our directory: sam_monarch_wing_20150415
Metadata for this directory:• Creator: Katherine McNeill• Subject: monarch butterfly wing• Description: this directory contains Sashimi ESEM images of a monarch
butterfly wing I took after finding a butterfly floating by the Charles River near MIT
• Contributor: Mark Clemente helped me with these images• Date: 20151015• Original Format: Sashimi Microscope format (.sam)• Relation: this is a directory that will contain multiple files• Type: image• Coverage: By the Charles River in
Cambridge, MA, MIT side• Rights: Monarch Butterfly Research
Foundation (funder) owns the data (grant number: 00213)
Metadata: our case study
Metadata for this image:• Title:
sam_monarch_wing_20150415_CM_001.tif
• Source: abcdefghijklmnopqrstuvwxyz.sam
• Relation: is a file in the directory: sam_monarch_wing_20150415
Metadata: our case study
In a filename
In a readme file
In a spreadsheet
In an XML file
Into a database
Metadata: capturing it
When choosing how to capture metadata, consider:• Expertise at your disposal• Complexity of your project• Collaborators• Your own comfort level
Active phase of research
Study Concept
Data Collection
Data Processing
Data Distribution
Data Analysis
Data Discovery
Data ArchivingRepurposing
Lifecycle adapted from DDI version 3.0 Combined Life Cycle Model
Storing data
What are your goals?
• To find a place to store your data?• To appropriately back up your data so that it’s not lost?
Data storage options at MIT
• IS&T options:• Athena lockers• Managed servers or Server co-location
• MIT research computing resources• DLC resources• Cloud storage (e.g., MIT Dropbox)
Backing up data: storing during active phase
Ideally, keep three copies of your data:
1. local/working copye.g. on your workstation or in shared workspace
2. remote copy e.g. on an MIT backup system
3. other copy in another remote location, or local/external e.g., external hard drive* (CDs and DVDs not built to last)
Test file recovery at setup and on a regular scheduleThis is not for publishing or sharing (storage ≠ archiving)
Backup services at MIT
• IS&T services: http://ist.mit.edu/backup• For laptops/desktops: CrashPlan PROe• For servers: Tivoli Storage Manager (TSM)• MIT Dropbox
Backing up data: storing during active phase
Additional considerations:
• What to keep? Consider weeding obsolete data as you go
• Huge file size!Images can often be very large files• Image compression may be done by your image software automatically • If not, you can compress your data.
• But always keep one uncompressed copy somewhere• Use open source compression software• Document the version of compression software used
Backing up data: storing during active phase
Additional considerations:
• My Research is top secret!• Then you can use encryption
• Don’t rely on 3rd party encryption alone• Use something like PGP (Pretty Good Privacy)
• Write the keys down on two pieces of paper• Store each piece of paper securely in separate locations
For more info. on this topic, see slides from our past workshop on Managing Confidential Data
File Organization &
File Formats
File organization: naming conventions
Naming conventions make life easier!
Naming conventions should be:• Descriptive• Consistent
Consider including:• Unique identifier (ie. Project Name or Grant # in folder name)• Project or research data name• Conditions (Lab instrument, Solvent, Temperature, etc.)• Run of experiment (sequential)• Date (in file properties too)• Version #
File organization: naming conventions
Naming conventions make life easier!
Naming conventions should be:• Descriptive• Consistent
YYYYMMDDMMDDYYYYYYMMDDMMDDYYMMDDDDMM
Sample001234Sample01234Sample1234
TimeDateDateProjectIDTimeProjectID
Maintain order
Include the same information
File organization: naming conventions
Best Practice ExampleLimit the file name to 32 characters (preferably less!) 32CharactersLooksExactlyLikeThis.csv
When using sequential numbering, use leading zeros to allow for multi-digit versions
For a sequence of 1-10: 01-10For a sequence of 1-100: 001-010-100
NO ProjID_1.csv ProjID_12.csv YES ProjID_01.csv ProjID_12.csv
Don’t use special characters& , * % # ; * ( ) ! @$ ^ ~ ' { } [ ] ? < > - NO name&date@location.doc
Use only one period and use it before the file extension
NO name.date.doc NO name_date..docYES name_date.doc
Avoid using generic data file names that may conflict when moved from one location to another
NO MyData.csvYES ProjID_date.csv
File organization: naming conventions
sam_monarch_wing_20150415_CM_001.tif
For our case study:
Sashimi Microscope format
Descriptive element
Date as YYYYMMDD
Initials because working in a group
Ascension # because part of a series
File format
Maybe started with:abcdefghijklmnopqrstuvwxyz.sam
File organization: file structure
sam_monarch_wing_20150415_CM_001.tif
For our case study:
sam_monarch_wing_20150415
File organization: versioning
Track versions of either:• Analysis/program/script files, while keep the original version of the data
file the same; OR• Data files themselves
File organization: versioning
In some cases, it may make sense to log the changes so that you can quickly assess and access the versions.
It’s good to document:• What was changed?• Who is responsible?• When did it happen?• Why?
CHANGELOG.md
CHANGELOG.txt
File formats
In the best case, your data files are both: • Non-proprietary (also known as open), and• Unencrypted and uncompressed
File formats: preferred examples
Proprietary Format Alternative/Preferred Format
Excel (.xls, .xlsx) Comma Separated Values (.csv)ASCII
Word (.doc, .docx)plain text (.txt), XML, PDF/A, HTML, ODF
or if formatting is needed, PDF/A (.pdf)
PowerPoint (.ppt, .pptx) PDF/A (.pdf), ODP, JPEG 2000, PDF, PNG
Photoshop (.psd) TIFF (.tif, .tiff), Quicktime (.mov) MPEG-4 (.mp4), MOV, AVI, MXFSounds WAVE, AIFFContainers TAR, GZIP, ZIPDatabases XML, CSV
File formats: converting considerations
Information can be lost when converting file formats.
To mitigate the risk of lost information when converting:– Note the conversion steps you take– If possible, keep the original file as well as the converted ones
File formats: our case study
• My data is in “Sashimi Environmental Scanning Electron Microscope (ESEM)” format!
– Problem: This is a proprietary format. How can I convert it to a more open-non proprietary format that will be usable in the future?
– Solution: I found Bio-Formats, a standalone Java library for reading and writing life sciences image file formats into alternative open formats.
Sashimi Format
OME-TIFFfile format
Proprietary format.sam
Open source format.tif
Sharing &
Preserving Data
Sharing and preserving data
Study Concept
Data Collection
Data Processing
Data Distribution
Data Analysis
Data Discovery
Data ArchivingRepurposing
Lifecycle adapted from DDI version 3.0 Combined Life Cycle Model
Sharing data: why?
• Further science as a whole• Further your research/reputation• Enable new discoveries with your data• Comply with funder/publisher data sharing requirements
Share Informally: Posting on a web site, sending via email upon request
Share Formally: Via a repository, which may also provide preservation and makes your data more accessible and citable
Preserving your data
What happens to your data when…
• the software you use to render it changes or becomes obsolete?
• the platform on which you manipulate it changes?• the hardware you created it on becomes obsolete?
Preserving your data
Preservation means storing your data in a place where it will be:• Stored• Backed up• Discoverable• Accessible for the future (as much as possible)
Preservation means that it is a particular person’s job to make every effort to make the data usable in the future.
Preservation = Long-term accessSome repositories ensure preservation of data over time
Storing data: repositories
Overall advantages:• May preserve your data for the future• Provides a metadata structure• Serves as a backup vehicle for your data• Makes sharing your data accessible and citable by others• May provide some computational/online analysis tools for
people to use your data• Gives your dataset a unique persistent identifier, e.g., DOI
As a researcher, you still need to:• Keep thorough documentation• Keep at least one copy of your data in an open, non-
proprietary format
Storing data: repository options
1. Domain repositories
Advantages:• Data is stored with similar datasets (by subject, format, or both)• Other researchers will find your data easily• The repository will understand what your data needs in terms of
storage, archiving, and preservation• Computational/online analysis tools may be available tailored to
analyzing that particular kind of data
Examples:• GenBank (for genome data)• ICPSR (for numeric social science data)
re3data.org | Registry of Research Data Repositories
Storing data: repository options
2. Institutional repositories
Advantages:• Linked to your institution• You can put all your datasets together (and collocate them with
publications)• University guarantees support of Institutional Repositories
• Some Domain Repositories may “go out of business” once their funding ends
Example:• DSpace@MIT
Storing data: our case study
Repository decision: Put our Monarch Wing Scan Data in DSpace@MIT
Reasoning:• Will be co-located with our other materials• TIFF format is supported (see DSpace policies)• No alternative domain repository
Additional considerations:• Must have faculty sponsorship• Each file cannot exceed 2.5 GB
Sharing Data: additional considerations
Intellectual Property & Confidentiality• Data is not copyrightable, but an expression of data can be.
• MIT or the funder may own your data (consult with the Technology Licensing Office)
• You can share your data if you, in fact, own it
• You can license data to limit what others can do with it (e.g., require attribution)
• It’s incumbent upon you to police usage of your data
• Look at Creative Commons Licenses, including the CC0 Declaration to emphatically put it in the public domain
• Confidential and sensitive data requires additional consideration and planning
Sharing Data: additional considerations
Journal requirements• Many journals require that underlying data accompany published
articles, usually found in “instructions for authors”
• More resources at http://libraries.mit.edu/data-management/share/journal-requirements/
Using other people’s data• Make sure that data doesn’t have a license agreement that prevents
you from sharing the data
• Most databases to which the MIT Libraries subscribe are licensed and carry restrictions on use, but many do allow for educational and research use, which allows for sharing limited portions of data.
Measure Twice, Cut Once
Research data management is an ongoing, but beneficial activity.
Things you want to check and re-check over time: • Is the data still stable and retrievable?• Is the metadata still available and understandable?• Are the formats still usable?
• Is the software still available?• Is any specialized hardware still available?
• Is the data still in the correct location?• Are my backups working as I expect?
Upcoming Data Management Workshops
• Research Funder Open Access Requirements from NASA, NSF, DOE and Other Federal Agencies, Wed, Jan 13, 2016, 14N-132
• Data Sharing and Storage, Thu, Jan 14, 1-2pm, 14N-132
• File Organization, Thu, Jan 21, 11am-12pm, 14N-132
• Data Management Plans and the DMPTool, Tue Jan 26, 11am-12pm, 14N-132
Contact: data-management@mit.edu
Questions? Comments? Tips?
Contact: data-management@mit.edu
Check out our web site: http://libraries.mit.edu/data-management
top related